Unravelling the distinct effects of VHL mutations and chromosome 3p loss in clear cell renal cell carcinoma: Implications for prognosis and treatment

Abstract

Dear Editor,

In this study, we delineated the distinct transcriptomic effects of VHL mutation and chromosome 3p (chr3p) loss, revealing that chr3p loss is specifically associated with immune suppression in clear cell renal cell carcinoma (ccRCC). Furthermore, we developed driver genomic aberration (DGA) gene signatures that demonstrate superior performance in predicting both patient prognosis and treatment response compared to traditional mutation-based approaches.

Renal cell carcinoma (RCC) accounts for 80%–85% of all primary kidney cancers, with ccRCC being the most common subtype (∼75%).¹ In 2023, ∼82 000 new RCC cases and ∼15 000 deaths were reported in the U.S. Despite surgery being curative for localised disease, ∼33% of patients relapse, and those with metastatic disease (∼15%) have a poor prognosis.¹ Despite treatment advances, significant variability in outcomes highlights the need for reliable molecular biomarkers to guide the treatment. Large-scale genomic studies such as the Cancer Genome Atlas (TCGA) have shown that the VHL gene is frequently inactivated in ccRCC through mutation or chr3p deletion.² However, the prognostic and therapeutic relevance of VHL mutations and chr3p loss remains controversial.

We examined the most frequently mutated genes in the TCGA KIRC dataset.² The VHL gene exhibited the highest mutation rate (52%), followed by PBRM1 (31%), SETD2 (11%) and BAP1 (5%) (Figure 1A). Copy number analysis revealed that chr3p loss occurred in 27% of patients, and most loss events, interestingly, encompassed these four genes. (Figure 1B). This genomic configuration is largely unique to ccRCC among TCGA cancer types (Figure S1A).

Mutation rates of these genes were consistent across tumour stages, indicating early tumourigenic roles (Figure S1B). To investigate the transcriptomic effects of VHL mutation and chr3p loss, we stratified TCGA-KIRC data by these aberrations and identified differentially expressed genes. In VHL-WT tumours, chr3p loss led to 1719 differentially expressed genes (DEGs; FDR < .05, |log₂FC| > 1.5), while in tumours with intact chr3p, VHL mutation resulted in 1577 DEGs. However, in the presence of VHL mutation, chr3p loss still induced 442 DEGs, whereas VHL mutation had no significant transcriptomic impact in chr3p-loss tumours, indicating that chr3p loss exerts a dominant regulatory effect (Figure 1C). Notably, immune-related genes were significantly enriched in genes that were downregulated in chr3p-loss tumours (Figure 1D), and GSEA analysis confirmed the suppression of immune pathways (Figure S2).

Given that immune gene suppression associated with chr3p loss, we explored its impact on the tumour immune microenvironment (TIME). Using previously reported data,³ we discovered that leukocyte and lymphocyte infiltration levels in TCGA-KIRC were significantly lower in chr3p-loss versus chr3p-WT tumours (p < .01), whereas VHL mutation had no significant impact on immune infiltration metrics (Figures 2A,B).

Using the TIMER algorithm,⁴ we estimated immune cell infiltration levels and observed that chr3p-loss tumours had significantly reduced infiltration of B cells, CD8+ and CD4+ T cells, macrophages, dendritic cells and neutrophils (Figure 2C). We also quantified BCR and TCR richness from RNA-seq reads and found both significantly diminished in chr3p-loss tumours, consistent with the immune suppression phenotype (Figures 2D,E). We next performed unsupervised clustering using expression of immune cell marker genes (Table S1) to classify tumours into immune “hot” and “cold” clusters. Chr3p-loss tumours were significantly enriched in the immune-cold cluster (79% vs. 41% in chr3p-WT; p = 9e-12, Fisher's exact test), while VHL-mutant tumours were evenly distributed, indicating that chr3p loss, not VHL, drives immune suppression (Figures 2F,G). To better understand this, we did stratified analysis, and chr3p loss alone showed reduced immune infiltration, which is not observed in VHL mutation only samples, suggesting a possible contribution from other genes from the region, such as BAP1 (Figure S3). Further supporting this, chr3p-loss samples exhibited significantly lower cancer-testis antigen (CTA) scores and reduced TGF-β pathway activity (Figure 2H). Both measures reflect diminished immune activity and T-cell function.

Despite the suppressed TIME caused by chr3p loss and the drastic transcriptomic change caused by the mutation, the prognostic significance and ability of the mutation status of chr3p loss and VHL remain contentious.^{5, 6} We therefore applied a transcriptomic signature-based approach by developing DGA gene signatures using TCGA mutation and copy number alteration (CNA) data to quantify the downstream effects of these alterations (Table S2). The VHL mutation signature achieved an AUC of 0.79 in classifying VHL mutation status and was validated in external datasets, including the Gordan cohort⁷ (Figure 3A) and CCLE data (Figure 3B), showing that the DGA signatures can capture genomic aberration dysregulated transcriptional activity.

We next evaluated the prognostic value of the DGA signatures. chr3p loss, VHL, PBRM1, SETD2, and BAP1 mutation derived signature is significantly associated with patient survival. Specifically, higher scores for PBRM1 and chr3p loss were linked to better prognosis, whereas elevated SETD2 and BAP1 signature scores correlated with poorer survival (Figure 3C). More specifically, while VHL mutation status alone did not show significant prognostic value (Figure 3D), the VHL signature score was strongly associated with improved overall survival in the TCGA cohort (Figure 3E), underscoring the utility of transcriptomic signatures in capturing functional pathway disruption beyond mutational status. This was replicated in the ICGC EU RCC cohort,⁸ where patients with high VHL scores had significantly longer survival (Figure 3F). As our VHL signature score reflects mutation-regulated transcriptomic dysregulation, it provides a continuous measure of such dysregulation. To further validate its prognostic utility, we performed a multivariate Cox regression survival analysis for both VHL wild-type and mutated patients (Figure 3G). The results demonstrate that the DGA signature scores remain significant predictors of patient outcomes, highlighting oncogenic pathway activities as main contributing factors to prognosis in ccRCC.

Given that sunitinib targets VEGFR-mediated angiogenesis, and VHL regulates hypoxia and angiogenesis, we investigated the correlation between VHL signature scores and tumour angiogenesis activity. Using an angiogenesis gene set, we found that angiogenic activity is significantly higher in VHL signature high samples (Figure 4A).

In the IMmotion150 trial,⁹ sunitinib responders had higher VHL and chr3p signature scores than non-responders (Figure 4B,C). Receiver operating characteristic analysis showed that the VHL signature predicted response with an AUC of 0.79, and the chr3p loss signature with an AUC of 0.68 (Figure 4D). These findings were validated in IMmotion151, further demonstrating the predictive capacity of the VHL signature for VEGF-targeted therapy response (Figure 4E).

To evaluate the utility of the chr3p loss signature in immunotherapy, we analysed patient outcomes in the IMmotion151 (combination therapy) and IMmotion150 (atezolizumab monotherapy) trials.⁹ Patients with higher chr3p scores were more likely to experience CR, PR, or SD, while low scores were associated with PD (Figure 4F). PD-L1 expression (Figure 4G) and tumour mutational burden (TMB, Figure S4) showed only weak associations with response. Interestingly, PD-L1 positivity inversely correlated with chr3p score (r = –0.31), suggesting that the chr3p signature scores provide independent insights (Figure 4H). Combining PD-L1 and chr3p signature score (top25% vs rest) improved response stratification: PD-L1-, chr3p-high patients showed lower response rates than all other groups (p = .03, Figure 4I). Multivariable Cox regression for progression-free survival (PFS) confirmed that the chr3p signature was the only significant predictor (HR = .85, p = .001), while PD-L1 was not (Figure 4J). High chr3p scores were associated with longer PFS in both IMmotion151 and IMmotion150 datasets (Figure 4K,L). In the latter, despite a smaller sample size, chr3p remained predictive (HR = .74, p = .02, Figure 4M). To further validate the result, we have applied it to CheckMate data and observed chr3p signature is protective regarding patient overall survival ¹⁰(Figure S5).

In summary, we demonstrate that chr3p loss, rather than VHL mutation, drives immune suppression in ccRCC. Although we have not experimentally validated the mechanism, the overall low genome instability of ccRCC and a lack of association between genome instability and immune response point to chr3p loss being the cause of such immune suppression. Using transcriptomic-based DGA signatures, we identify distinct impacts of these alterations on prognosis and treatment response. The VHL and chr3p signatures outperform mutation status in predicting patient survival and guiding target and immunotherapy. Although our results are validated across independent clinical datasets, we acknowledge the absence of experimental validation and prospective trials as a limitation due to the scope of this paper. The scores can be applied to the clinical setting with a threshold simply using zero or better stratification based on the patient population in clinical testing. Our findings provide a clinically relevant framework for stratifying ccRCC patients and advancing precision oncology and guiding clinical trial design by paring immune checkpoint therapy with VEGFR inhibitors.

Cheng Chao conceived the project. Cheng Chao and Xiang Wang obtained the data. Xiang Wang and Cheng Chao developed the methods. Xiang Wang and Cheng Chao performed computational analyses. Xiang Wang and Cheng Chao wrote the manuscript. Xiang Wang and Cheng Chao interpreted the results. Xiang Wang and Jian-Rong Li made figures. Cheng Chao supervised the project. All authors critically reviewed the content. All authors read and approved the final manuscript.

The authors declare no conflict of interest.

This work is supported by the Cancer Prevention Research Institute of Texas (CPRIT) (RR180061) and the National Cancer Institute of the National Institutes of Health (1R01CA269764).

Not applicable.